The seismic models of three-dimensional (3-D) mantle heterogeneity may be interpreted in terms of the density perturbations which drive mantle flow and thus provide important predictions of flow-related observables, such as the nonhydrostatic geoid. The current models of global-scale shear velocity heterogeneity provide reasonably good fits to the long-wavelength nonhydrostatic geoid data (60--70% variance reductions) but rather poor fits to the corresponding free-air gravity anomalies (30--40% variance reductions). This major difference is due to two factors: (1) the very different amplitude spectrum of the gravity anomalies, which is nearly flat for degrees l3.

The largest mismatch between the pattern of predicted and observed gravity anomalies is in the southern hemisphere. This observation suggests that one reason for the poor overall match between the two fields may be that the global seismic data are not resolving sufficiently well the heterogeneity in the southern hemisphere portion of the deep mantle. In contrast, the gravity data provide accurate and geographically uniform constraints on the vertically integrated heterogeneity in the mantle. We therefore perform a series of experiments in which we simultaneously invert a large set of seismic data (which includes long-period waveforms, SS-S and ScS-S differential travel times, and normal-mode structure coefficients) and the long-wavelength gravity anomaly data. We thus determine whether it is possible to derive new 3-D heterogeneity models which satisfy both data sets. The gravity anomaly data are interpreted in the context of spherically-symmetric viscous flow models of the mantle. In these inversion experiments we test several radial viscosity and Δln&rgr;/Δlnv profiles and thereby assess their plausibility.

The joint seismic-geodynamic inversions reveal that it is indeed possible to greatly improve the fit to the free-air gravity anomalies (with variance reductions of 80--90% readily accessible) while preserving the fit to the seismic data. This improvement is achieved with some adjustment to the heterogeneity in the depth range 1500--2500 km, where the seismic data constraints appear to be weakest. The joint inversions also reveal new structures in the southern hemisphere portion of the lower mantle which apparently are not resolved by the seismic data alone. ¿ American Geophysical Union 1994